AI questions worth bringing to your rav
Most of the difficulty is not halachic, it is that the question is hard to ask clearly. Here is how to describe what the technology does, so the sheilah is the real one.
To be unambiguous at the outset: nothing in this article is a psak, and nothing here should be read as one. These are questions, not answers, and the answers belong to your rav.
What this can usefully do is something else. A great deal of the difficulty in asking a sheilah about AI is that the technology has to be described accurately before the question makes sense, and most descriptions are either too vague to answer or wrong in ways that change the answer. So below is a plain description of what the thing does, followed by the questions that actually arise.
Describing it accurately
A language model is not a search engine and it is not a person. It does not look anything up, and it does not know anything in the ordinary sense. It predicts likely continuations of text, having been trained on an enormous quantity of writing.
Three consequences matter for almost every sheilah that follows:
- It produces fluent, confident text whether or not it is correct, and it cannot tell you which it has done.
- It has no source. When it answers a halachic question it is not citing a sefer; it is producing text that resembles the way such answers are written.
- It is not consistent. The same question can produce different answers, so testing it once tells you very little.
The questions that come up
On Torah content
- May it be used to look up a halacha, given that it does not cite a source and may be confidently wrong?
- Is there a difference between using it to find a source you then check, and relying on its answer?
- May it be used to help prepare a shiur or a dvar Torah, and does authorship need to be disclosed?
- What is the status of Torah text it generates — is there a concern about how it is printed, stored or discarded?
On children and chinuch
- At what age, if any, and with what supervision?
- If a filtered tool exists, does that change the answer relative to an unfiltered one?
- What should a parent do having discovered a child has been using an unfiltered tool?
On work
- May it be used to write material for a client that the client believes you wrote?
- Is there an issue with pasting in a third party's confidential information?
- Are generated images of people a tznius question, a different question, or no question?
On Shabbos
- Scheduling a task before Shabbos that completes during it — the ordinary questions about amirah and about benefit.
- A tool that continues to run on a server over Shabbos with no interaction from you.
What makes the conversation easier
Two practical things. First, bring a specific case rather than the general topic — "my son wants to use this for a history project" is answerable in a way that "what about AI" is not. Second, be able to say plainly whether what you are using filters its content, whether that filtering is yours to set, and whether anyone can see what was asked. Those facts change some answers, and a rav should not have to guess at them.
It is also worth saying that these questions are new, that thoughtful people are working through them, and that an answer today may be refined later. That is not a reason to avoid asking.
Common questions
Can AI answer halachic questions?
It will produce a fluent answer, which is not the same thing. A language model does not cite a source and cannot tell you whether it is correct. Treat anything it produces on halacha as a starting point to check, never as an answer, and bring the actual sheilah to your rav.
What should I be able to tell my rav about the tool we use?
Whether it filters prompts, answers or both; whether the filtering is set by you or by the vendor; whether anyone can see what was asked; and whether it covers images as well as text. Those facts can change the answer.
Is this article a psak?
No. It is a list of questions and a plain description of how the technology works, so the sheilah you bring is the real one. The answers belong to your rav.
Published by Navōn. How these guides are written and checked.
Read next
- What makes an AI kosher?
There is no hechsher for software. So what should a frum school or family actually check before letting students use an AI tool? A practical standard.
- Scheduling AI availability around Shabbos and Yom Tov
The practical side of making an AI tool unavailable over Shabbos and Yom Tov — what a scheduling feature has to get right, and what belongs to your rav.
- Talking to your child about AI
What to say at different ages, the questions children actually ask, and why the goal is being the person they tell rather than the person who checks.